How To Create a Synthetic Data Set Like Charlie Javice

There’s something funny about how legacy financial firms react when a tech founder under 35 walks in wearing a hoodie, armed with a pitch deck and a buzzword or two. The old guard melts. They lean in. They don’t just open the checkbook — they hand over the whole account.

Charlie Javice knew that. So did Sam Bankman-Fried. So did Elizabeth Holmes. These weren’t isolated scams. They were carefully rehearsed performances for an audience desperate to believe in youth, innovation, and disruption — no matter how synthetic the story might be.

In Charlie’s case, the story was about student data. She was the founder of Frank, a startup supposedly helping students apply for financial aid more easily. According to prosecutors, she told JPMorgan she had over 4 million users. In reality, she had around 300,000. The rest? Generated. Faked. Fabricated in a data set designed to look real enough to pass a billion-dollar sniff test.

And it worked — until it didn’t.

This article isn’t about glorifying that move. It’s about understanding how synthetic data works, why this keeps happening, and what it says about the state of old money chasing millennial clout.

What Is Synthetic Data, Really?

Synthetic data is fake information that mimics real information. If you’ve ever used a test account on a website, or seen a demo with “John Smith” as the user, you’ve seen it in action.

In the right hands, it’s useful — for training AI models, for building product prototypes, or for privacy-preserving analytics. In the wrong hands? It’s a way to invent users, transactions, or entire markets out of thin air.

In Charlie’s case, prosecutors allege she hired a data science professor to help her create a list of fake students. Fake names, fake emails, fake schools, fake financial aid status. All formatted in a way that looked like it came straight out of a legitimate CRM.

Something like:

Anna Jackson at anna.jackson@collegeplanner, zip code 10012, attending NYU, FAFSA status complete.
Marcus Lin at…